关键词: 航空设备故障, 知识图谱, 注意力机制, 双向门控循环神经网络, 故障诊断

Abstract: In response to the challenge of large volumes of aviation equipment failure data and the inefficiency of traditional fault diagnosis methods, we have employed knowledge graph technology to build a high-performance graph database. This database replaces the conventional relational databases used in civilian aircraft to enhance the efficiency of fault diagnosis decision-making. Additionally, we have optimized and improved the knowledge extraction model through the integration of an attention mechanism and Bidirectional Gated Recurrent Unit (BiGRU). Initially, we designed the ontology of the knowledge graph based on expert experience, clearly defining entities and relationship types within the knowledge graph. Subsequently, we trained the BiGRU-Attention optimized knowledge extraction model using fault-text corpora annotated with BIO tags. This training aims to enhance the efficiency of extracting entities and relationships from unstructured text. Comparisons with classical knowledge extraction models reveal that the BiGRU-Attention improved model demonstrates superior recognition performance. Ultimately, we utilized the extracted entities and relationships to construct a knowledge graph for diagnosing faults in aviation equipment. This knowledge graph facilitates more accurate fault diagnosis for maintenance personnel involved in aviation equipment troubleshooting.

Key words: aviation equipment malfunction, knowledge graph, attention mechanism, bidirectional gated recurrent neural network, fault diagnosis